Sven Mika
0de41e4a6b
[RLlib] Trainer sub-class QMIX/MAML/MB-MPO (instead of build_trainer). ( #20639 )
2021-12-02 13:17:10 +01:00
Sven Mika
3d2e27485b
[RLlib] Trainer sub-class DQN/SimpleQ/APEX-DQN/R2D2 (instead of using build_trainer). ( #20633 )
2021-11-30 18:05:44 +01:00
Artur Niederfahrenhorst
d07e50e957
[RLlib] Replay buffer API (cleanups; docstrings; renames; move into rllib/execution/buffers dir) ( #20552 )
2021-11-19 11:57:37 +01:00
gjoliver
d81885c1f1
[RLlib] Fix all the CI tests that were broken by is_training and replay buffer changes; re-comment-in the failing RLlib tests ( #19809 )
...
* Fix DDPG, since it is based on GenericOffPolicyTrainer.
* Fix QMix, SAC, and MADDPA too.
* Undo QMix change.
* Fix DQN input batch type. Always use SampleBatch.
* apex ddpg should not use replay_buffer_config yet.
* Make eager tf policy to use SampleBatch.
* lint
* LINT.
* Re-enable RLlib broken tests to make sure things work ok now.
* fixes.
Co-authored-by: sven1977 <svenmika1977@gmail.com>
2021-10-28 18:06:47 +02:00
gjoliver
99a0088233
[RLlib] Unify the way we create local replay buffer for all agents ( #19627 )
...
* [RLlib] Unify the way we create and use LocalReplayBuffer for all the agents.
This change
1. Get rid of the try...except clause when we call execution_plan(),
and get rid of the Deprecation warning as a result.
2. Fix the execution_plan() call in Trainer._try_recover() too.
3. Most importantly, makes it much easier to create and use different types
of local replay buffers for all our agents.
E.g., allow us to easily create a reservoir sampling replay buffer for
APPO agent for Riot in the near future.
* Introduce explicit configuration for replay buffer types.
* Fix is_training key error.
* actually deprecate buffer_size field.
2021-10-26 20:56:02 +02:00
gjoliver
89fbfc00f8
[RLlib] Some minor cleanups (buffer buffer_size -> capacity and others). ( #19623 )
2021-10-25 09:42:39 +02:00
Sven Mika
ed85f59194
[RLlib] Unify all RLlib Trainer.train() -> results[info][learner][policy ID][learner_stats] and add structure tests. ( #18879 )
2021-09-30 16:39:05 +02:00
Sven Mika
9c9b482661
[RLlib] Allow n-step > 1 and prio. replay for R2D2 and RNNSAC. ( #18939 )
2021-09-29 21:31:34 +02:00
Sven Mika
4888d7c9af
[RLlib] Replay buffers: Add config option to store contents in checkpoints. ( #17999 )
2021-08-31 12:21:49 +02:00
Thomas Lecat
c02f91fa2d
[RLlib] Ape-X doesn't take the value of prioritized_replay into account ( #17541 )
2021-08-16 22:18:08 +02:00
Sven Mika
5a313ba3d6
[RLlib] Refactor: All tf static graph code should reside inside Policy class. ( #17169 )
2021-07-20 14:58:13 -04:00
Sven Mika
1fd0eb805e
[RLlib] Redo fix bug normalize vs unsquash actions (original PR made log-likelihood test flakey). ( #17014 )
2021-07-13 14:01:30 -04:00
Amog Kamsetty
bc33dc7e96
Revert "[RLlib] Fix bug in policy.py: normalize_actions=True has to call unsquash_action, not normalize_action." ( #17002 )
...
This reverts commit 7862dd64ea
2021-07-12 11:09:14 -07:00
Sven Mika
7862dd64ea
[RLlib] Fix bug in policy.py: normalize_actions=True has to call unsquash_action, not normalize_action. ( #16774 )
2021-07-08 17:31:34 +02:00
Sven Mika
7318439c3d
[RLlib] DQN native_ratio (for training intensity) incorrect (discussion 1763). ( #15436 )
...
Thanks @Manuscrit !
2021-04-22 11:06:29 +02:00
Sven Mika
4f66309e19
[RLlib] Redo issue 14533 tf enable eager exec ( #14984 )
2021-03-29 20:07:44 +02:00
SangBin Cho
fa5f961d5e
Revert "[RLlib] Issue 14533: tf.enable_eager_execution() must be called at beginning. ( #14737 )" ( #14918 )
...
This reverts commit 3e389d5812
2021-03-25 00:42:01 -07:00
Sven Mika
3e389d5812
[RLlib] Issue 14533: tf.enable_eager_execution() must be called at beginning. ( #14737 )
2021-03-24 12:54:27 +01:00
Sven Mika
732197e23a
[RLlib] Multi-GPU for tf-DQN/PG/A2C. ( #13393 )
2021-03-08 15:41:27 +01:00
Sven Mika
8000258333
[RLlib] R2D2 Implementation. ( #13933 )
2021-02-25 12:18:11 +01:00
Sven Mika
19c8033df2
[RLlib] Fix most remaining RLlib algos for running with trajectory view API. ( #12366 )
...
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* LINT and fixes.
MB-MPO and MAML not working yet.
* wip
* update
* update
* rmeove
* remove dep
* higher
* Update requirements_rllib.txt
* Update requirements_rllib.txt
* relpos
* no mbmpo
Co-authored-by: Eric Liang <ekhliang@gmail.com>
2020-12-01 17:41:10 -08:00
Sven Mika
b6b54f1c81
[RLlib] Trajectory view API: enable by default for SAC, DDPG, DQN, SimpleQ ( #11827 )
2020-11-16 10:54:35 -08:00
Sumanth Ratna
9da7bdcc8e
Use master for links to docs in source ( #10866 )
2020-09-19 00:30:45 -07:00
desktable
4ccfd07a61
[RLlib] Add docstrings for agents/dqn ( #10710 )
2020-09-15 12:37:07 +02:00
desktable
799318d7d7
[RLlib] Add type annotations for agents/dqn ( #10626 )
2020-09-09 18:55:26 +02:00
Sven Mika
28ab797cf5
[RLlib] Deprecate old classes, methods, functions, config keys (in prep for RLlib 1.0). ( #10544 )
2020-09-06 10:58:00 +02:00
Sven Mika
78dfed2683
[RLlib] Issue 8384: QMIX doesn't learn anything. ( #9527 )
2020-07-17 12:14:34 +02:00
Piotr Januszewski
155cc81e40
Clarify training intensity configuration docstring ( #9244 ) ( #9306 )
2020-07-05 20:07:27 -07:00
Eric Liang
34bae27ac7
[rllib] Flexible multi-agent replay modes and replay_sequence_length ( #8893 )
2020-06-12 20:17:27 -07:00
Sven Mika
2746fc0476
[RLlib] Auto-framework, retire use_pytorch in favor of framework=... ( #8520 )
2020-05-27 16:19:13 +02:00
Eric Liang
9a83908c46
[rllib] Deprecate policy optimizers ( #8345 )
2020-05-21 10:16:18 -07:00
Eric Liang
aa7a58e92f
[rllib] Support training intensity for dqn / apex ( #8396 )
2020-05-20 11:22:30 -07:00
Eric Liang
2c599dbf05
[rllib] Port QMIX, MADDPG to new execution API ( #8344 )
2020-05-07 23:41:10 -07:00
Eric Liang
b14cc16616
[rllib] Enable functional execution workflow API by default ( #8221 )
2020-05-05 12:36:42 -07:00
Eric Liang
2298f6fb40
[rllib] Port DQN/Ape-X to training workflow api ( #8077 )
2020-04-23 12:39:19 -07:00
Sven Mika
428516056a
[RLlib] SAC Torch (incl. Atari learning) ( #7984 )
...
* Policy-classes cleanup and torch/tf unification.
- Make Policy abstract.
- Add `action_dist` to call to `extra_action_out_fn` (necessary for PPO torch).
- Move some methods and vars to base Policy
(from TFPolicy): num_state_tensors, ACTION_PROB, ACTION_LOGP and some more.
* Fix `clip_action` import from Policy (should probably be moved into utils altogether).
* - Move `is_recurrent()` and `num_state_tensors()` into TFPolicy (from DynamicTFPolicy).
- Add config to all Policy c'tor calls (as 3rd arg after obs and action spaces).
* Add `config` to c'tor call to TFPolicy.
* Add missing `config` to c'tor call to TFPolicy in marvil_policy.py.
* Fix test_rollout_worker.py::MockPolicy and BadPolicy classes (Policy base class is now abstract).
* Fix LINT errors in Policy classes.
* Implement StatefulPolicy abstract methods in test cases: test_multi_agent_env.py.
* policy.py LINT errors.
* Create a simple TestPolicy to sub-class from when testing Policies (reduces code in some test cases).
* policy.py
- Remove abstractmethod from `apply_gradients` and `compute_gradients` (these are not required iff `learn_on_batch` implemented).
- Fix docstring of `num_state_tensors`.
* Make QMIX torch Policy a child of TorchPolicy (instead of Policy).
* QMixPolicy add empty implementations of abstract Policy methods.
* Store Policy's config in self.config in base Policy c'tor.
* - Make only compute_actions in base Policy's an abstractmethod and provide pass
implementation to all other methods if not defined.
- Fix state_batches=None (most Policies don't have internal states).
* Cartpole tf learning.
* Cartpole tf AND torch learning (in ~ same ts).
* Cartpole tf AND torch learning (in ~ same ts). 2
* Cartpole tf (torch syntax-broken) learning (in ~ same ts). 3
* Cartpole tf AND torch learning (in ~ same ts). 4
* Cartpole tf AND torch learning (in ~ same ts). 5
* Cartpole tf AND torch learning (in ~ same ts). 6
* Cartpole tf AND torch learning (in ~ same ts). Pendulum tf learning.
* WIP.
* WIP.
* SAC torch learning Pendulum.
* WIP.
* SAC torch and tf learning Pendulum and Cartpole after cleanup.
* WIP.
* LINT.
* LINT.
* SAC: Move policy.target_model to policy.device as well.
* Fixes and cleanup.
* Fix data-format of tf keras Conv2d layers (broken for some tf-versions which have data_format="channels_first" as default).
* Fixes and LINT.
* Fixes and LINT.
* Fix and LINT.
* WIP.
* Test fixes and LINT.
* Fixes and LINT.
Co-authored-by: Sven Mika <sven@Svens-MacBook-Pro.local>
2020-04-15 13:25:16 +02:00
Eric Liang
31b40b00f6
[rllib] Pull out experimental dsl into rllib.execution module, add initial unit tests ( #7958 )
2020-04-10 00:56:08 -07:00
Sven Mika
22ccc43670
[RLlib] DQN torch version. ( #7597 )
...
* Fix.
* Rollback.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* Fix.
* Fix.
* Fix.
* Fix.
* Fix.
* WIP.
* WIP.
* Fix.
* Test case fixes.
* Test case fixes and LINT.
* Test case fixes and LINT.
* Rollback.
* WIP.
* WIP.
* Test case fixes.
* Fix.
* Fix.
* Fix.
* Add regression test for DQN w/ param noise.
* Fixes and LINT.
* Fixes and LINT.
* Fixes and LINT.
* Fixes and LINT.
* Fixes and LINT.
* Comment
* Regression test case.
* WIP.
* WIP.
* LINT.
* LINT.
* WIP.
* Fix.
* Fix.
* Fix.
* LINT.
* Fix (SAC does currently not support eager).
* Fix.
* WIP.
* LINT.
* Update rllib/evaluation/sampler.py
Co-Authored-By: Eric Liang <ekhliang@gmail.com>
* Update rllib/evaluation/sampler.py
Co-Authored-By: Eric Liang <ekhliang@gmail.com>
* Update rllib/utils/exploration/exploration.py
Co-Authored-By: Eric Liang <ekhliang@gmail.com>
* Update rllib/utils/exploration/exploration.py
Co-Authored-By: Eric Liang <ekhliang@gmail.com>
* WIP.
* WIP.
* Fix.
* LINT.
* LINT.
* Fix and LINT.
* WIP.
* WIP.
* WIP.
* WIP.
* Fix.
* LINT.
* Fix.
* Fix and LINT.
* Update rllib/utils/exploration/exploration.py
* Update rllib/policy/dynamic_tf_policy.py
Co-Authored-By: Eric Liang <ekhliang@gmail.com>
* Update rllib/policy/dynamic_tf_policy.py
Co-Authored-By: Eric Liang <ekhliang@gmail.com>
* Update rllib/policy/dynamic_tf_policy.py
Co-Authored-By: Eric Liang <ekhliang@gmail.com>
* Fixes.
* WIP.
* LINT.
* Fixes and LINT.
* LINT and fixes.
* LINT.
* Move action_dist back into torch extra_action_out_fn and LINT.
* Working SimpleQ learning cartpole on both torch AND tf.
* Working Rainbow learning cartpole on tf.
* Working Rainbow learning cartpole on tf.
* WIP.
* LINT.
* LINT.
* Update docs and add torch to APEX test.
* LINT.
* Fix.
* LINT.
* Fix.
* Fix.
* Fix and docstrings.
* Fix broken RLlib tests in master.
* Split BAZEL learning tests into cartpole and pendulum (reached the 60min barrier).
* Fix error_outputs option in BAZEL for RLlib regression tests.
* Fix.
* Tune param-noise tests.
* LINT.
* Fix.
* Fix.
* test
* test
* test
* Fix.
* Fix.
* WIP.
* WIP.
* WIP.
* WIP.
* LINT.
* WIP.
Co-authored-by: Eric Liang <ekhliang@gmail.com>
2020-04-06 11:56:16 -07:00
Sven Mika
5537fe13b0
[RLlib] Exploration API: ParamNoise Integration into DQN; working example/test cases. ( #7814 )
2020-04-03 10:44:25 -07:00
Eric Liang
9392cdbf74
[rllib] Add high-performance external application connector ( #7641 )
2020-03-20 12:43:57 -07:00
Eric Liang
dd70720578
[rllib] Rename sample_batch_size => rollout_fragment_length ( #7503 )
...
* bulk rename
* deprecation warn
* update doc
* update fig
* line length
* rename
* make pytest comptaible
* fix test
* fi sys
* rename
* wip
* fix more
* lint
* update svg
* comments
* lint
* fix use of batch steps
2020-03-14 12:05:04 -07:00
Eric Liang
f5d12a958b
[rllib] Port Ape-X to distributed execution API ( #7497 )
2020-03-12 00:54:08 -07:00
Eric Liang
a644060daa
[rllib] First pass at pipeline implementation of DQN ( #7433 )
...
* wip iters
* add test
* speed up
* update docs
* document it
* support serial sampling
* add test
* spacing
* annotate it
* update
* rename to pipeline
* comment
* iter2 wip
* update
* update
* context test
* update
* fix
* fix
* a3c pipeline
* doc
* update
* move timer
* comment
* add piepline test
* fix
* clean up
* document
* iter s
* wip dqn
* wip
* wip
* metrics
* metrics rename
* metrics ctx
* wip
* constants
* add todo
* suppport .union
* wip
* support union
* remove prints
* add todo
* remove auto timer
* fix up
* fix pipeline test
* typing
* fix breakage
* remove bad assert
* wip
* fix multiagent example
* fixapply
* update a3c
* remove a2c pl
* 0 workers
* wip
* wip
* share metrics
* wip
* wip
* doc
* fix weight sync and global var updates
* mode
* fix
* fix
* doc
* fix
2020-03-07 14:47:58 -08:00
Sven Mika
510c850651
[RLlib] SAC add discrete action support. ( #7320 )
...
* Exploration API (+EpsilonGreedy sub-class).
* Exploration API (+EpsilonGreedy sub-class).
* Cleanup/LINT.
* Add `deterministic` to generic Trainer config (NOTE: this is still ignored by most Agents).
* Add `error` option to deprecation_warning().
* WIP.
* Bug fix: Get exploration-info for tf framework.
Bug fix: Properly deprecate some DQN config keys.
* WIP.
* LINT.
* WIP.
* Split PerWorkerEpsilonGreedy out of EpsilonGreedy.
Docstrings.
* Fix bug in sampler.py in case Policy has self.exploration = None
* Update rllib/agents/dqn/dqn.py
Co-Authored-By: Eric Liang <ekhliang@gmail.com>
* WIP.
* Update rllib/agents/trainer.py
Co-Authored-By: Eric Liang <ekhliang@gmail.com>
* WIP.
* Change requests.
* LINT
* In tune/utils/util.py::deep_update() Only keep deep_updat'ing if both original and value are dicts. If value is not a dict, set
* Completely obsolete syn_replay_optimizer.py's parameters schedule_max_timesteps AND beta_annealing_fraction (replaced with prioritized_replay_beta_annealing_timesteps).
* Update rllib/evaluation/worker_set.py
Co-Authored-By: Eric Liang <ekhliang@gmail.com>
* Review fixes.
* Fix default value for DQN's exploration spec.
* LINT
* Fix recursion bug (wrong parent c'tor).
* Do not pass timestep to get_exploration_info.
* Update tf_policy.py
* Fix some remaining issues with test cases and remove more deprecated DQN/APEX exploration configs.
* Bug fix tf-action-dist
* DDPG incompatibility bug fix with new DQN exploration handling (which is imported by DDPG).
* Switch off exploration when getting action probs from off-policy-estimator's policy.
* LINT
* Fix test_checkpoint_restore.py.
* Deprecate all SAC exploration (unused) configs.
* Properly use `model.last_output()` everywhere. Instead of `model._last_output`.
* WIP.
* Take out set_epsilon from multi-agent-env test (not needed, decays anyway).
* WIP.
* Trigger re-test (flaky checkpoint-restore test).
* WIP.
* WIP.
* Add test case for deterministic action sampling in PPO.
* bug fix.
* Added deterministic test cases for different Agents.
* Fix problem with TupleActions in dynamic-tf-policy.
* Separate supported_spaces tests so they can be run separately for easier debugging.
* LINT.
* Fix autoregressive_action_dist.py test case.
* Re-test.
* Fix.
* Remove duplicate py_test rule from bazel.
* LINT.
* WIP.
* WIP.
* SAC fix.
* SAC fix.
* WIP.
* WIP.
* WIP.
* FIX 2 examples tests.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* Fix.
* LINT.
* Renamed test file.
* WIP.
* Add unittest.main.
* Make action_dist_class mandatory.
* fix
* FIX.
* WIP.
* WIP.
* Fix.
* Fix.
* Fix explorations test case (contextlib cannot find its own nullcontext??).
* Force torch to be installed for QMIX.
* LINT.
* Fix determine_tests_to_run.py.
* Fix determine_tests_to_run.py.
* WIP
* Add Random exploration component to tests (fixed issue with "static-graph randomness" via py_function).
* Add Random exploration component to tests (fixed issue with "static-graph randomness" via py_function).
* Rename some stuff.
* Rename some stuff.
* WIP.
* update.
* WIP.
* Gumbel Softmax Dist.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* WIP
* WIP.
* WIP.
* Hypertune.
* Hypertune.
* Hypertune.
* Lock-in.
* Cleanup.
* LINT.
* Fix.
* Update rllib/policy/eager_tf_policy.py
Co-Authored-By: Kristian Hartikainen <kristian.hartikainen@gmail.com>
* Update rllib/agents/sac/sac_policy.py
Co-Authored-By: Kristian Hartikainen <kristian.hartikainen@gmail.com>
* Update rllib/agents/sac/sac_policy.py
Co-Authored-By: Kristian Hartikainen <kristian.hartikainen@gmail.com>
* Update rllib/models/tf/tf_action_dist.py
Co-Authored-By: Kristian Hartikainen <kristian.hartikainen@gmail.com>
* Update rllib/models/tf/tf_action_dist.py
Co-Authored-By: Kristian Hartikainen <kristian.hartikainen@gmail.com>
* Fix items from review comments.
* Add dm_tree to RLlib dependencies.
* Add dm_tree to RLlib dependencies.
* Fix DQN test cases ((Torch)Categorical).
* Fix wrong pip install.
Co-authored-by: Eric Liang <ekhliang@gmail.com>
Co-authored-by: Kristian Hartikainen <kristian.hartikainen@gmail.com>
2020-03-06 10:37:12 -08:00
Sven Mika
83e06cd30a
[RLlib] DDPG refactor and Exploration API action noise classes. ( #7314 )
...
* WIP.
* WIP.
* WIP.
* WIP.
* WIP.
* Fix
* WIP.
* Add TD3 quick Pendulum regresison.
* Cleanup.
* Fix.
* LINT.
* Fix.
* Sort quick_learning test cases, add TD3.
* Sort quick_learning test cases, add TD3.
* Revert test_checkpoint_restore.py (debugging) changes.
* Fix old soft_q settings in documentation and test configs.
* More doc fixes.
* Fix test case.
* Fix test case.
* Lower test load.
* WIP.
2020-03-01 11:53:35 -08:00
Sven Mika
d537e9f0d8
[RLlib] Exploration API: merge deterministic flag with exploration classes (SoftQ and StochasticSampling). ( #7155 )
2020-02-19 12:18:45 -08:00
Sven Mika
6e1c3ea824
[RLlib] Exploration API (+EpsilonGreedy sub-class). ( #6974 )
2020-02-10 15:22:07 -08:00
Sven Mika
2ccf08ad10
[RLlib] Bug fix: DQN goes into negative epsilon values after reaching explora… ( #6971 )
...
* Bug fix: DQN goes into negative epsilon values after reaching exploration percentage.
* Add `epsilon_initial_eps` to SAC to pass test_nested_spaces.py.
* Add `exploration_initial_eps` to QMIX default config.
2020-01-31 09:54:12 -08:00
Sven Mika
e6227082bd
[RLlib] Add torch flag to train.py ( #6807 )
2020-01-17 18:48:44 -08:00
Sven
60d4d5e1aa
Remove future imports ( #6724 )
...
* Remove all __future__ imports from RLlib.
* Remove (object) again from tf_run_builder.py::TFRunBuilder.
* Fix 2xLINT warnings.
* Fix broken appo_policy import (must be appo_tf_policy)
* Remove future imports from all other ray files (not just RLlib).
* Remove future imports from all other ray files (not just RLlib).
* Remove future import blocks that contain `unicode_literals` as well.
Revert appo_tf_policy.py to appo_policy.py (belongs to another PR).
* Add two empty lines before Schedule class.
* Put back __future__ imports into determine_tests_to_run.py. Fails otherwise on a py2/print related error.
2020-01-09 00:15:48 -08:00
